Some of America’s largest technology companies are calling on the US government to make open artificial intelligence a national priority as China continues to build on its lead in the field.
Their message reveals gaps in U.S. policy. The US has a chip strategy for AI. We need an open model strategy now.
Proponents argue that an open model can reduce costs, keep sensitive data within a company’s own systems, and prevent companies from relying on a small number of AI providers. This model allows more researchers and security teams to examine strong models and find weaknesses.
Unlike ChatGPT and Claude, customers access it through a service controlled by. OpenAI and humanopenweight models can be downloaded, customized, and run on a company’s own computers. This results in lower prices and gives users more control over their data and technology.
Last week, a coalition led by Nvidia Published an open letter arguing that U.S. AI leadership depends on building a strong open ecosystem. Microsoft, Meta, and other major technology companies have signed on. OpenAI and Google later added support.
A shocking cybersecurity incident that occurred a few days ago showed why this issue is important.
During internal testing, OpenAI models found a way out of restricted environments and compromised systems on the AI platform Hugging Face. When Hug Face tried to investigate, it was unable to analyze the attack due to undisclosed safety controls in the commercially available model. The company instead used an open Chinese model that allowed it to run on its own infrastructure.
It was an undisclosed American model that caused the incident. China’s open model helped in that investigation.
This episode complicates the argument that closed AI is always safe.
The United States has restricted China’s access to advanced chips, spent billions of dollars on land-based semiconductor manufacturing, and encouraged a massive expansion of data centers and power. On the other hand, Chinese research institutes Chipu Moonshot AI has released a capable open model at a fraction of the cost of leading US systems.
The Trump administration is reportedly currently considering restrictions on Chinese-made models. But even a ban won’t stop these models from spreading around the world. It may simply put American developers on the sidelines.
China has clear incentives to promote open weight AI. Each improvement makes more powerful models cheaper, weakening the advantage of American companies that charge a premium for access to closed systems.
OpenAI and Anthropic have opposite incentives. Their business depends on keeping their best models exclusive. Anthropic CEO Dario Amodei also warned that models that are available for anyone to download are difficult to monitor and can be modified by malicious actors.
These security concerns are real, but so is the risk of allowing China to become the foundation upon which the rest of the world is built.
The way to win the open source race is to outbuild, not ban.
The United States could support open AI in the same way it supported chips by giving universities and startups access to computing power, awarding government contracts to U.S. open model developers, and funding the security tools needed to run those models securely.
Businesses are already looking for alternatives, and open models reduce costs and allow businesses to keep sensitive data within their own systems.
“The data these companies have is truly a strategic asset,” said Vipul Ved Prakash, CEO of AI platform Together AI. He said companies risk giving away their “business recipe” by sending it to manufacturers with strong closed models.
Adoption is progressing rapidly. In the last week of June, Chinese models accounted for 48% of the traffic tracked by OpenRouter, up from 20% a year earlier. The US model dropped from 74% to 32%.
The future of AI may not be determined by who builds the single smartest model. It may depend on who builds the model for others to run.
Now it’s more and more China.
